The FRONTIER Act, introduced on July 23, 2026, by Representatives Jay Obernolte and Lori Trahan, aims to establish a unified federal regulatory framework for advanced AI models. The bill, designated H.R. 9925, seeks to replace fragmented state-level rules with a consistent national standard.
The legislation targets large developers with annual revenue exceeding $50 million and AI development spending surpassing $1 billion over a 36-month period. A frontier model is defined as any system requiring more than 10^26 compute operations for training.
Companies meeting these criteria must publish safety frameworks, perform annual audits, and report critical incidents. The largest developers are required to engage licensed independent verification organizations for external assessments, a provision recently supported by OpenAI.
The bill proposes creating the position of Under Secretary for AI Security within the Department of Commerce. This official would oversee compliance and incident tracking, with the Commerce Secretary gaining emergency authority to restrict deployments in cases of catastrophic risk.
Noncompliance carries significant financial consequences, with civil penalties potentially reaching $10 million per day. The final version of the bill includes exemptions for state authority regarding deployment-specific matters and child safety.
